PromptPilot: task scheduler for Claude Code and Codex that works while you sleep

Token quotas, request limits, reset windows — all dictated by the provider, not the developer. But developers retain one resource under personal control: the timing of task submission. That's exactly what PromptPilot does — a prompt queue scheduler for AI CLI tools that lets you launch a code agent without being at your monitor.

What is PromptPilot and why is it needed?

PromptPilot is an open-source tool written by one developer for themselves, but published with the understanding that others face the same situation. The scenario is simple: you want to run a task on Claude Code or Codex at midnight — when the provider resets its limit — but you're already asleep. Or conversely: in the morning you open your laptop and want the session warm-up to have already happened without you.

The system accepts tasks from three interfaces: a standard terminal, a web interface, and a Telegram bot. A worker picks them from the queue and executes them sequentially. If the tool hits a rate limit — it doesn't crash, but pauses and retries automatically. This is especially crucial for those working with free or limited tiers, where every limit reset requires manual restart.

Architecture: SQLite and three processes

The technical implementation is intentionally minimalist. The entire queue is stored in SQLite — without external brokers like Redis or RabbitMQ. Three processes: API server, worker, and Telegram bot — communicate through the database. This allows everything to run on a regular laptop or inexpensive VPS without complex infrastructure.

Tasks have priorities and execution times. You can set a task with the label "run at 03:00" — the worker will pick it up at the right moment. Claude Code, Codex, and Qwen are supported — essentially any CLI tool that accepts prompts via standard input. Tasks with higher priority will be executed before others, even if queued later.

Telegram bot as a control panel

A separate scenario — interactive interaction with the model through the bot. If the AI agent asks a clarifying question mid-task, the user gets a notification in Telegram and can answer right there, without opening the terminal. This is critical if the task was submitted before leaving the house or the agent is stuck waiting for input while you slept.

The bot doesn't just show task status, but acts as a full two-way communication channel with the model. For those who often work on the move or from mobile, this is a significant detail: you don't lose control of the agent even if you're physically far from your computer.

Why this matters

PromptPilot is not a product from a large company and not a SaaS with a paid subscription. It's a tool built for a specific pain point: the mismatch between a developer's work schedule and the provider's rate limit schedule. The author admits the project was made for themselves — but that's precisely why it solves a real, not invented, problem.

As AI CLI tools become a standard part of the workflow, the need to manage task flow and limits will only grow. PromptPilot closes this gap with simple means — no Kubernetes, no message queues, no complex deployment. One SQLite, three processes, a Telegram bot — and your AI agent works on schedule. The project code is published, deployment is described in the README. This is a case where open-source solves exactly the problem that commercial tools don't notice — because it's too small to build a separate product around.